Lausanne 2011
Inequality and economic
crises
A B Atkinson, Nuffield College, Oxford
Salvatore Morelli, University of Oxford
1
1. Introduction;
2. Economic crises 1911-2010;
3. Which inequality of what?
4. Empirical evidence: case studies of Nordic and Asian
crises;
5. Do crises lead to inequality?
6. Does inequality lead to crises?
7. What can we conclude so far?
2
1. Introduction: social cohesion and economic
crises
Two way relationship:
Inequality and lack of social cohesion
Economic crises: financial crises
and
collapses in output/consumption
3
DIFFERENT VIEWS:
How do crises affect inequality?
FALL: US 1929 Great Crash: “The upward drift [in inequality]
accelerates from the turn of the century up to America’s
entrance into World War I. Inequality fell between 1929 and
the early years after World War II” (Williamson and Lindert,
1980, page 95).
“The share of years … that a country was exposed to a
banking crisis has a substantive negative impact on top
income shares” [5 year crisis reduces share of top 1 per cent
by 1 percentage point]” (Roine, Vlachos and Waldenström,
Journal of Public Economics, 2009).
4
How do crises affect inequality?
RISE: Asian financial crisis of 1997: “After nearly a decade of
either declining or stable trend since the mid 1980s, the
family income inequality in Korea sharply increased in the
course of the financial crisis, and remained high even after
the economy recovered from the recession” (Lee, 2002).
“The current economic crisis has shown that it is the poor
and vulnerable groups in society who are disproportionately
affected by such shocks” (OECD, January 2011).
5
DIFFERENT VIEWS:
Does inequality increase the risk of crises?
NO: The indexes to three authoritative accounts of financial
crises, by Kindleberger and Aliber (2005), Krugman (2009)
and Reinhart and Rogoff (2009), contain neither “inequality”
nor “income distribution”.
The US Financial Crisis Inquiry Commission, set up in 2009 to
investigate “the most significant financial crisis since the
Great Depression”, was charged with examining 22 specific
areas. None of these refer to inequality.
6
Does inequality increase the risk of crises?
YES: According to Stiglitz, in the face of stagnating real
incomes, households in the lower part of the distribution in
the US borrowed to maintain a rising standard of living. This
borrowing later proved unsustainable, leading to default and
pressure on over-extended financial institutions.
According to Rajan, “growing income inequality in the
United States stemming from unequal access to quality
education led to political pressure for more housing credit.
This pressure created a serious fault line that distorted
lending in the financial sector.”
7
2. Economic crises 1911-2010
Consider:
• systemic banking crises (not limited to a few banks);
• “collapses” in real consumption per capita.
The study of crises requires long run data: “a data set that
covers only twenty-five years simply cannot give one an
adequate perspective” (Reinhart and Rogoff, 2009).
It requires cross-country data: “to use history to gauge the
probability and size distribution of macroeconomic disasters,
it is hopeless to rely on the experience of a single country”
(Barro, 2009, page 246).
8
The data challenge: Banking crises
We have relied on three major sources to identify systemic
banking crises:
• Bordo, Eichengreen, Klingebiel and Martinez-Peria, 2001;
• Reinhart and Rogoff, 2008, 2009 and Reinhart, 2010;
• Laeven and Valencia, 2009 and 2010.
They do not cover all the same countries or the same time
periods, and they do not always agree.
We have applied a majoritarian criterion. Where there are
only two entries (one data-base does not cover the country
or period), and they disagree, we have in general included
the crisis.
9
Figure 1 62 Banking crises in 25 countries over 100 years (exc war times)
US
Norway
Sweden
Finland
Iceland
India
Japan
Indonesia
Malaysia
Mauritius
Singapore
Argentina
Brazil
Australia
Canada
New Zealand
South Africa
France
Germany
Italy
Netherlands
Portugal
Spain
Switzerland
UK
1911
1921
1931
1941
1951
1961
1971
1981
1991
2001
10
The data challenge: Consumption crises
Barro defines consumption (or GDP) “disaster” as peak to
trough decline of at least 10 per cent: e.g. consumption in
Argentina fell from a peak in 1998 to a trough in 2002 by
22.5 per cent.
On the basis that perception of a “crisis” depends on
expectations, we apply the Barro criterion before 1950, but
a cut-off of 5 per cent after 1950.
11
Figure 2 55 consumption "collapses" in 25 countries over 100 years (exc war)
US
Norway
Sweden
Finland
Iceland
India
Japan
Indonesia
Malaysia
Mauritius
Singapore
Argentina
Brazil
Australia
Canada
New Zealand
South Africa
France
Germany
Italy
Netherlands
Portugal
Spain
Switzerland
UK
1911
1921
1931
1941
1951
1961
1971
1981
1991
2001
12
Systemic No systemic TOTAL
banking banking
crisis
crisis
Consumption
18
37
55
“collapse”
No
consumption
“collapse”
44
TOTAL
62
13
First need to clarify
3. Which inequality of what?
• Inequality of what? Earnings versus income versus
consumption versus wealth;
• Snapshots versus lifetime outcomes; inequality of
opportunity;
• Which part of the parade should we be watching?
• Horizontal dimensions of inequality.
14
Chartbook of economic
inequality: 5 indicators
3. Poverty
rate
Poverty
line
1. Overall
inequality:
Gini
coefficient
Income
2. Top
income
share
“Middle
class”
The income “parade”
+ (4) top earnings decile and (5) top wealth share
15
Inequality: the data challenge
• Crises are rare events, so that we need a long run of years;
• To explore the impact of a crisis, we need to be able to
monitor change year by year: we need annual series;
• For the present crisis, we lack up-to-date distributional data
for many countries;
• For past years, we cannot simply download annual series on
inequality covering a range of countries;
• Data have to be pieced together from a variety of national
sources; data for earlier parts of the century are hard (or
impossible) to find;
• Priority given to time series consistency over cross-country
comparability.
16
Figure 3 Banking crises for which distributional data
US
Norway
Nordic crises
Sweden
Finland
Iceland
India
Asian crises
Japan
Indonesia
Malaysia
Mauritius
Singapore
Argentina
Brazil
Australia
62 banking crises:
distributional data
for 35.
Canada
New Zealand
South Africa
France
Germany
Italy
Netherlands
Portugal
Spain
Switzerland
UK
1911
1921
1931
1941
1951
1961
1971
1981
1991
2001
17
Figure 4 consumption "collapses" for which distributional data
US
Norway
Sweden
Finland
Iceland
India
Japan
Indonesia
Malaysia
Mauritius
Singapore
Argentina
Brazil
Australia
Canada
New Zealand
South Africa
55 consumption
collapses:
distributional data
for 33
France
Germany
Italy
Netherlands
Portugal
Spain
Switzerland
UK
1911
1921
1931
1941
1951
1961
1971
1981
1991
2001
18
4. Empirical evidence: case studies of Nordic
and Asian crises
Figure NO1 Economic crises and inequality in Norway 1911-2010
40
170
35
160
30
150
Gini coefficient, equivalised (EU-scale) household income,
weighted by persons
Share of top 1 per cent in gross income
20
Per cent living in households with equivalised (EU-scale)
disposable income below 60 per cent median
Share of top 1 per cent in total wealth
140
Per cent
Per cent
25
Earnings at top decile as % median, series 1 (RH scale)
130
Earnings at top decile as % median, series 2 (RH scale)
15
120
10
110
5
0
1911
100
1921
1931
1941
1951
1961
1971
1981
1991
2001
Vertical line indicates start of banking crises; rectangle shows consumption collapse (peak to trough)
19
Figure SWE1 Economic crises and inequality in Sweden 1911-2010
Gini coefficient, equiv after tax income using EU scale household
income, weighted by persons
Share of top 1 per cent in gross income
35
175
Share of top 1 per cent in equiv after tax income using EU scale
household income, weighted by persons
Per cent below 60 per cent median
30
150
Top earnings decile as % median (RH scale)
25
125
20
100
15
75
10
50
5
25
0
1911
0
1921
1931
1941
1951
1961
1971
1981
1991
Per cent
Per cent
Share of top 1 per cent in total wealth (RH scale)
2001
Vertical line indicates start of banking crisis; rectangle shows consumption collapse (peak to trough)
20
Figure FIN1 Economic crises and inequality in Finland 1911-2010
40
Income Distribution Survey, equiv after tax income using EU scale
household income, weighted by persons
200
Share of top 1 per cent in gross income, series 1
190
Share of top 1 per cent in gross income, series 2
180
Per cent below 60 per cent of median
30
Top decile of earnings (RH scale)
170
Per cent
25
160
20
150
Per cent
35
140
15
130
10
120
5
110
0
1911
100
1921
1931
1941
1951
1961
1971
1981
1991
2001
Vertical line indicates start of banking crisis; rectangle shows consumption collapse (peak to trough)
21
Figure JA1 Economic crises and inequality in Japan
200
Gini coefficient, Income Redistribution
Survey
70
Share of top 1 per cent in gross income
190
Share of top 0.1 per cent in gross income
60
180
Per cent
50
Wealth Gini coefficient
170
Earnings top decile as per cent of median
(RH scale)
160
40
150
30
Per cent
Per cent below 60% median
140
130
20
120
10
110
0
100
1911
1921
1931
1941
1951
1961
1971
1981
1991
2001
Vertical line indicates start of banking crisis
22
60
24
50
20
40
16
30
12
Per cent
Per cent
Figure INDON1 Economic crises and inequality in Indonesia 1911-2010
Gini, expenditure data
Gini, income data
20
8
Income share top 0.05 per cent (RH scale)
Per cent below poverty line
Income share top 1 per cent (RH scale)
10
4
0
0
1911
1921
1931
1941
1951
1961
1971
1981
1991
2001
Vertical lines indicate start ofbanking crises; rectangles show consumption collapses (peak to trough)
23
Figure MYA1 Economic crises and inequality in Malaysia 1911-2010
30
60
Income share top 1 per cent
Income share top 0.1 per cent
25
50
Share of bottom 40 per cent
Per cent
Gini gross income (RH scale)
20
40
15
30
10
20
5
10
0
1911
0
1921
1931
1941
1951
1961
1971
1981
1991
2001
Vertical lines indicate start of banking crises; rectangles show consumption collapses (peak to trough)
24
Figure SI1 Economic crises and inequality in Singapore 1911-2010
60
250
Gini coefficient among employed population,
series 1
Gini coefficient among employed households,
income from work after government benefits and
taxes, series 3
Share of top 1 per cent in gross income
40
Per cent
225
Gini coefficient among households, ranked by
income from work, series 2
200
Share of top 10 per cent in gross income
30
175
Per cent
50
Earnings at upper quintile as % median (RH scale)
20
150
10
125
0
100
1911
1921
1931
1941
1951
1961
1971
1981
1991
2001
Vertical line indicates start of banking crisis; rectangle shows consumption collapse (peak to trough)
25
5. Do crises lead to inequality?
Figure US1929 Window diagram
Figure US1984-88 Window diagram
4.0
2.0
4.0
12.00
3.0
9.00
2.0
6.00
1.0
3.00
0.0
0.00
-1.0
-3.00
0.0
-2.0
-4.0
Gini coefficient
-2.0
Gini coefficient
-3.0
Income share top 1 per cent, disposable income (exc capital
gains)
Income share top 1 per cent, disposable income (inc capital -9.00
gains)
Percent in poverty
-4.0
Share of top 1 per cent in total wealth
-6.0
Income share top 1 per cent, disposable
income (exc capital gains)
Income share top 1 per cent, disposable
income (inc capital gains)
Share of top 1 per cent in total wealth
-8.0
-6.00
-12.00
Top decile as % median (Right hand scale)
-5.0
-10.0
t-5
t-4
t-3
t-2
t-1
t
t+1
t+2
t+3
t+4
t+5
-15.00
t-5
t-4
t-3
t-2
t-1
t
t+1
t+2
t+3
t+4
t+5
Figure US2007 Window diagram
Window
diagrams
4.0
8.0
2.0
4.0
0.0
0.0
-2.0
-4.0
Gini coefficient
-4.0
-8.0
Income share top 1 per cent, disposable
income (exc capital gains)
Income share top 1 per cent, disposable
income (inc capital gains)
Percent in poverty
-6.0
-12.0
Top decile as % median (Right hand scale)
-8.0
t-5
t-4
t-3
t-2
t-1
t
t+1
t+2
t+3
t+4
t+5
-16.0
?
26
Evidence from all 25 countries 1911-2010
Did inequality rise before and fall after? Classification of 31 banking crises (4?).
After
\
Classic
US
1929
Before
TOTAL
=
/
4
6
4
14
3
3
6
12
\
1
2
2
5
TOTAL
8
11
12
31
/
=
27
Evidence from all 25 countries 1911-2010
Did inequality rise before and fall after? Classification of 33 consumption
collapses.
After
\
Classic
US
1929
Before
/
=
\
TOTAL
TOTAL
=
/
6
3
4
13
3
1
6
10
4
2
2
8
13
6
12
31
28
6. Does inequality lead to crises? Evidence
from all countries 1911-2010
Level of inequality in 2007 compared with ten years earlier and
identification of a banking crisis in 2007-8
GINI coefficient
Identified
crisis
No
identified
crisis
TOTAL
Higher
inequality
2
5
7
No higher
inequality
4
10
14
TOTAL
6
15
21
29
A matter of judgment:
• classification of banking crises (B);
• classification of consumption collapses (B/A);
• identification of direction of change in inequality (C/B).
The data do not lend themselves to straightforward
statistical tests.
30
Interpretation: Co-incident or causal?
1. Banking model, with competitive consumption:
Increased demand for consumer borrowing to finance desired consumption to
keep up with those whose earnings are rising faster; banks respond by raising
rates but take on more risk. Change in inequality (top, overall and bottom) is
causal.
2. Banking model, with introduction of securitization:
Change in banking practices with introduction of securitization, taking on
greater risk to an extent that is greater the higher the degree of inequality.
Level of inequality (overall and bottom) is jointly causal.
3. Banking model, with shift in remuneration practices:
Remuneration becomes tied more closely to sales, so that banks behave more
like sales maximisers than maximisers of shareholder value, increasing the
exposure to risk. Observe increased top inequality and increased risk of crisis.
Co-incident, not causal.
31
4. Financial sector model, with bubbles:
Asset bubble draws skilled workers into financial sector, causing wage
dispersion to rise. Change in inequality (top) is co-incident, not causal.
5. Political economy model of monetary policy:
In response to rise in inequality, uses deregulation of banking for
distributional reasons. Change in inequality (overall and bottom) is causal.
6. Political economy model of deregulation:
Increased inequality at the top leads to lobbying for deregulation. Change in
inequality (top) is co-incident.
7. Political economy model of pensions:
Government decides to reduce size of welfare state. Loss of income to
current beneficiaries causes inequality to rise. Households respond by saving
more in private pensions, driving up equity prices, and by “buy-to-let”
purchases of housing, driving up house prices. Change in inequality (bottom)
is co-incident, not causal.
32
7. What can we conclude?
• Economic inequality has many dimensions; here focused on income and
its components, but some of the most important dimensions of inequality
from the standpoint of social cohesion may be those not measured, such
as inequality of opportunity;
• Heterogeneity is important; different parts of the distribution may
change differently: it depends which part of the parade we are watching;
different parts are relevant to different explanatory models and have
differing impact on social cohesion;
• The role of inequality in the origins of crises and the distributional
impact of banking crises may differ over time: “this time it may be
different”; specifically, in the US there was a substantial rise in inequality
leading up to the 1929 and S+L crises, but this was not the case with the
present crisis; on the other hand, in terms of levels of inequality, 2007
may be more like 1929 than the 1980s;
33
7. What can we conclude (continued)
• Outside the US, the history of crises in different countries round the
world does not suggest that either rising or high levels of inequality have
been adduced as significant causal factors; there is a range of possible
mechanisms, but it is not evident that there is a smoking gun;
• Cannot write off high inequality as temporary feature of bubble; in the
US the only sustained period of inequality-reduction was in the early
1940s; quite a number of European (and Asian) countries have seen
inequality and poverty rise after a banking crisis;
• Much unfinished business, and, in order to make progress, need to bring
together different branches of social science.
34